Fault diagnosis method of wind power transmission system based on depth generation adversarial network

Abstract

The invention discloses a fault diagnosis method of a wind power transmission system based on a depth generation adversarial network. The fault diagnosis method of the wind power transmission system based on the depth generation adversarial network comprises the following steps: existing wind turbine generator transmission system vibration data of different kinds of working conditions under load are collected, and large wind turbine generator transmission system cross-domain sample training set is established; depth generation adversarial network modules are constructed; the depth generation adversarial network is pre-trained; and an on-line diagnosis to the large wind turbine generator transmission system is carried out by adopting trained modules. According to the fault diagnosis methodof the wind power transmission system based on the depth generation adversarial network, according to similarity and difference of data between a source domain and a target domain, the depth generation adversarial network is used to melt and rectify data between the two domains, similarity features can be extracted from cross domain data by a multi-layer stack self-coding network structure, and differences between the source domain and the target domain can be further rectified by a domain discriminator based on a softmax sorter, and a malfunction diagnosis of the target domain can be realizedby adopting rich knowledge of the source domain.

Description

technical field [0001] The invention relates to the field of wind power, in particular to a fault diagnosis method for a wind power transmission system based on a deep generative confrontation network. Background technique [0002] The wind turbine transmission system is one of the most important parts of the entire wind turbine generator. Due to the long-term light-heavy load, high-low speed and harsh external environment, the wind turbine transmission system is prone to failure. At this time, a stable, Intelligent fault diagnosis method is particularly important. [0003] In the era of the Internet of Things and Industry 4.0, bearing health monitoring systems collect a large amount of real-time data, enabling artificial intelligence methods to effectively mine features and diagnose faults, such as support vector machines (SVM), artificial neural networks (ANN). However, the effectiveness of these methods requires that the original training data set and the target data set...

AI Technical Summary

Problems solved by technology

However, the effectiveness of these methods requires that the original training data set and the target data set obey the same distribution, which is difficult to meet in the actual working environment of the system. Therefore, fault diagnosis methods based on domain adaptation have been developed, which mainly use The similarity between multiple domains extracts cross-domain common features, thus enabling the use of source domain data to solve the fault diagnosis problem of target domain data, but traditional domain adaptation methods also have two main shortcomings: 1) When the target domain When the data are not sufficiently representative, domain adaptation methods may become less effective for defining domain difference distances

2) Shallow structures tend to ignore the useful knowledge shared by multiple sources, and it is difficult to adapt the knowledge to the target domain during the testing phase

Images